Integrated Device Testing
Audio Subsystem Testing
Audio Playback Testing
Using a selection of Hard Rock and Heavy Metal music tracks and Windows Media Player, the audio subsystem playback performance was tested for playback accuracy and fidelity.
Playback using the audio test tracks was distortion free using a 5.1 speaker setup through the on-board analogue audio ports.
Testing was also performed using the rear panel headphone port (lower right port in rear panel audio port block) using a Razer Carcharias audio headset to test the audio subsystem's gain switching technology. Using the selected audio tracks, audio reproduction was much cleaner with the gain switch set to position 2, resulting in a 6.5x gain. With the gain switch set properly, the audio clarity experience during the listening session resulted in some of the best audio I have heard from an on-board audio solution.
Microphone Port Testing
For testing the board's Microphone input port, the microphone from a Razer Carcharias audio headset was used to capture a 30 second spoken phrase with the assistance of the Microsoft Sound Recorder application. The resulting audio file was saved to the desktop and played back using Windows Media Player.
The recorded audio test was distortion-free and without any detected aberrant noise effects with the recording-related settings dialed in correctly. We did not need to enable Microphone Boost, but found that pickup was best with input volume set at a minimum of 80. Further if Microphone Boost was enabled or Acoustic Audio Cancellation was disabled (via the SoundBlaster control app), minor audio feedback was heard real-time and in the recorded audio.
ATTO Disk Benchmark
To validate that the board’s device ports were functioning correctly, we connected an OCZ Vertex 3 90GB SATA III SSD to the system and ran the ATTO Disk Benchmark against the drive. The SSD was directly connected to the native SATA 3 ports as well as the SATA 3 ports associated with the integrated SATA Express ports. ATTO was configured to test against transfer sizes from 0.5 to 8192 KB with Total Length set to 512 MB and Queue Depth set to 10. The SSD selected for testing has a maximum read throughput of 550 MB/s and a write throughput of 500 MB/s on a SATA III controller. The drive tests were repeated three times with the highest repeatable read and write speeds recorded.
Drive performance on the Intel Z97 controller was significantly better than that on the Marvell controller. There was no difference in drive performance on the Intel Z97 controller when connecting to SATA or SATA Express tied device ports. In either case, drive performance pushed the limits of the device's rated specs. The Marvell drive performance fell well short of that seen on the Intel Z97 controller with its device read speeds coming in at just under 400 MB/s but is write speeds barely breaking 300 MB/s.
SoftPerfect Research NetWorx Speed Test
In conjunction with Windows Performance Monitor, SoftPerfect Research NetWorx Speed Meter application was used to measure the upload and download performance of the motherboards integrated network controllers. Speed Meter was used to measure average network throughput in MB/s with Windows Performance Monitor used to measure average CPU utilization during the tests.
The LanBench network benchmarking software was used to generate send and receive traffic between the local and remote systems over a five minute period with packet size set to 4096 and connection count set to 20. A LanBench server was set up on the remote system to generate or receive traffic for the tests performed. The upload and download tests were repeated three times with the highest repeatable average throughput and the lowest repeatable average CPU utilization percentage recorded.
Note that that theoretical maximum throughput for a Gigabit Ethernet adapter is 125 MB/s (1.0 Gbps).
Both the Intel and Killer NIC controllers performed on par with one another with boht upload and download speeds averaging close to 120 MB/s. The Intel controller's advantage was with its CPU utilization while in use. The CPU usage remained under 10% during transfer testing with the Intel controller while averaging 10-20% when transferring data with the Killer-based controller.